Neural Networks and NLP Strategies - Part 3

Neural Networks and NLP Strategies - Part 3

The NLP Strategies Model

By Joseph O’Connor and Brian Van der Horst

This is the third in our series of articles, updating the NLP model of strategies with recent advances in neural network computing and cognitive science. This article will review the traditional NLP thinking about strategies, link back to the programming metaphor in NLP, and move from there towards a richer strategic model.

Modelling

Strategies are a small part of the broader picture of NLP modelling. NLP started as a goal oriented technology of achievement . Competence was only the beginning, it dealt in the currency of excellence. Historically, the first models were Gregory Bateson, Virginia Satir, Fritz Perls and Milton Erickson. From Gregory Bateson came systemic thinking and epistemology. From Virginia Satir and Fritz Perls came representational systems, the therapeutic application of parts of the Meta Model, and reframing. Milton Erickson balanced what was until then a rather conscious approach, with his work on eliciting unconscious resources, use of metaphor, and what was to become known as the Milton Model of language and the art of anchoring. The first models were framed mostly as communication skills, and psychotherapeutic procedures. From these, NLP grew to become a systematic approach to the structure of our subjective experience. NLP has since grown by modelling excellence in education, sport, and business.

Modelling is the process of discovering how a person gets the results they get. How they think, the representational systems, metaprograms and logical levels they employ. What language patterns they use and the physiology they adopt. NLP has grown by modelling performers in a field, but the same tools can be used to model how people become depressed or create lives that are unhappy and unfulfilled. Knowing then how they do it, you can often intervene effectively by educating them how to stop themselves, and/or how to generate more desirable and compelling behaviors.

Modelling is very pragmatic. A model is an overall map of a given phenomenon. It should also contain explicit information of what steps should be taken in what order to obtain a given outcome.

A model cannot be proved true or false, only that it works or not.

Musical notation is often given as a simile for NLP modelling:

if two different musicians on opposite sides of the globe can produce roughly similar results on different instruments, it is because their sheet music is an effective model of the musical composition.

When you can model another's skill and get the same results as your model, you can then devise a means to teach it to others. This is the basis of effective NLP training.

Modelling in any field produces new techniques and skills, and further tools for modelling. NLP modelling is generative because the results can be applied to the next modelling process to make it even more effective. Modelling is the basis of accelerated learning.

One problem in modelling is the “barrier of consciousness’. People have minimal awareness of how they do what they do, and some, such as Virginia Satir do not want to know either, fearing the knowledge would destroy their innocence and therefore their ability, like dreams fading when exposed to the light of day.

Modelling a skill

To model a skill, you focus on what the model does (behaviour and physiology), how he does it, (internal thinking strategies) and why he does it (the supporting beliefs and values).

To model a skill you need:

The behaviour and physiology.

The internal thinking processes.

The beliefs and values.

NLP modelling has given rise to some exaggerated claims. For example: possible for one, so possible for all, which would make an excellent political slogan if NLP were ever to go in that direction. This comes again from the Artificial Intelligence metaphor, where programs can be transferred from computer to computer without change and run in exactly the same way. Human beings are somewhat more complicated. You cannot take a skill out of context without changing the skill. If you model Einstein’s thought process, you will not become another Einstein, however you will be able to enrich and broaden your own thinking using his strategies.

Strategies

Here is a very brief resume of traditional NLP thinking about strategies. A strategy is the internal thought process of the model: a sequence of representations the person goes through in order to achieve their outcome. A strategy has three crucial aspects:

* The representation systems used.

* The distinctions within the representation systems (submodalities)

* The sequence of the steps.

There are some principles of well formed strategies too: it should have an outcome, and it should involve all three main representational systems (Visual, Auditory and Kinesthetic). There should be no loops without an exit point and there should be at least one step that relates to the outside world.

Representation Systems

In a strategy, the senses may be turned towards the external world or used internally as representation systems.

First the visual system (V). We can see events in the external world (Ve), and we can visualise internally (Vi). The pictures we make mentally may be remembered (Vir), from something we have originally seen in the outside world, or they may be constructed (Vic). For example when we read novels for enjoyment, we will usually construct pictures in our imagination of the scenes described in the book. Technical reading is more difficult to remember, because it is harder to construct pictures of abstract concepts.

The auditory system (A) divides into six aspects. We can hear sounds from the external world (Ae) or from our internal world (Ai). These sound are a combination of tonality (At), and digital: the words we hear, regardless of tonality(Ad). When we hear sounds internally, (Ai) they may be remembered (Air) or constructed, (Aic). Talking to ourselves is usually referred to as internal dialogue or auditory internal digital (Aid), although the voices we use in our head have some tonality, and sometimes they do not even belong to us.

The kinesthetic sense is more mixed and controversial. There are sensations from the outside world (Ke). The internal sensations ( Ki) can be divided broadly into emotions, remembered sensations and feelings of balance and bodily awareness, often described as the vestibular system. We taste with the gustatory sense (G) and smell with the olfactory (O), both of which can be external or internal.

Submodalities

Submodalities are the distinctions we make within a representational system. For instance, brightness, distance and motion. Some, such as direction, are common to all representational systems. Submodalities may be discontinuous, that is on or off, for example associated or dissociated. A memory cannot be both at once. Or they may be continuous or analogic, varying on a sliding scale, such as brightness or volume.

Sometimes submodalities are a key element of how effective a strategy will be. For example a soft soothing, internal voice, and some slowly moving, dark pictures work well as a going to sleep strategy. A loud internal voice and big, fast moving, bright pictures are excellent as an insomnia strategy. The submodalities themselves are neither good nor bad, it all depends what you want to do.

The more submodality distinctions you can make and the finer you make them, the more choices you have about strategies. These distinctions may have to do with where you have your thresholds. We set thresholds and adjust them continually so the brain can pay attention to what contrasts and comparisons are important for a given outcome. Below a certain threshold the data will not register. This is true of brains and neural network computers. The degree of sensory acuity we have to stimuli from the outside world sets the threshold below which the sensation will not be perceived and therefore stored as what Stephen Grossberg calls an instar pattern in long term memory. It therefore cannot form part of a comparison between long term and short term memory. For example a musician critically listens to sounds and makes fine distinctions in pitch, rhythm and tonality. These fine distinctions are present and can be used to evaluate a musical interpretation, by comparing what they hear (Ae) with what they want to hear (Aic). The NLP idea of ‘preferred representational system’ may be a mixture of inbuilt and trained thresholds.

Richard Bandler has said that submodalities are the little arrows and bars in strategies: they are how human beings progress from one step to another in their thinking processes.

Sequence

Strategies are also often described with a cooking metaphor. First you need the right ingredients: the representational systems. You need the right qualities and quantities of the ingredients: the submodalities, and you must do things in the right order. Adding the eggs after you have put the cake in the oven gets a very different result from adding them to the mixture before.

Sequence brings out the greatest differences between the old linear model of strategies and what we are proposing in terms of neural networks. The traditional strategic model operating in straight lines is a simplified, useful fiction. It is the sort of model an Artificial Intelligence computer would produce and not code congruent with the way the brain works.

Language itself also limits how we think about models. Words follow one after the other in time and matters it much very order which go in they. The Meta Model pattern of cause-effect is built into the American-English language. Nevertheless we make a multitude of multi layered meanings from every sentence and metaphors express many things simultaneously.

Motivation strategy

Take an example of a simple motivation strategy. One person may first look at some work that needs to be done (Ve). They hear an internal voice (Aid) that says something like “Time to do this” in a fairly loud, encouraging tonality. They then construct a mental picture (Vic) of the work when it is done in comparison to how it currently looks (Vir). They make this picture big and bright and shiny, (critical submodalities). They feel good looking at that picture (Ki+). Then they start the work. This strategy works well and is pleasant to run. It runs on a metaprogram of moving towards the positive.

We could notate this as : Ve--->Aid--->Vic/Vir--->Ki+

Here is a second example of a motivation strategy. This person looks at the work to be done (Ve). They hear a nagging internal voice (Aid) that says, “Hey, you really should get on with this.” They feel resistant (Ki-). They make a picture of everything else they would rather do (Vic), different from the work that is to be done (Vir) and feels better (Ki+). Then, as a polarity response (p), they make a picture of the consequences of not doing the work, and feel bad (Ki-) . This may go on as a loop between negative kinesthetic and the internal voice for a number of days (or months) until the picture of the undesirable consequences gets very close and large, and the consequent feeling becomes very uncomfortable. When this threshold is reached, they do the task. This strategy gets the task done in the end, as long as the uncomfortable feeling hits the threshold in time, but is less pleasant. It runs on a metaprogram of avoiding the negative.

In notation: Ve--->Aid--->Ki---->Vic/Vir--->K+--p-->Vic--->Ki-

Structurally, this gives us a model of how someone motivates himself easily, and how another procrastinates. Obviously, if we want to transfer either of these competencies, we would have to take copious notes of the sub-modalities, or as many people are doing now, using their own short-hand for the key submodality distinctions.

The Five Fundamental Strategies

In standard NLP training, there are structurally five basic strategies taught, because all macro-strategies of complex behaviours can be chunked down to these five, it is worth a little review.

1. Memory. People need have a way of encoding and retrieving experience. This is the strategy we are talking about when we talk about the Lead System and the Primary Reresentational System. Put otherwise, how to file experience and later, how to open up and understand your representations of experience. The lead is analogous to DOS system in a personal computer, the prime rep is much like if you use a program like Wordstar or Word.

2. Belief. We have to have a way of testing between the constructed representations of our imagination and our remember “real” memories. This is often referred to as a reality, convincer strategy, and was orginally taught as the Reference System.

3. Decision. This is where human being begin to factor in the criteria produced by the second strategy. When you make evaluations you decide. Life is a long river of little decisions. This is the strategy at the heart of the TOTE model.

4. Motivation. Here we often use a piece of our memory, a chunk of imagination, and of course, a decision piece. But some people do this so idiosyncratically, that we tend to treat it as a separate genre, as in the two examples above.

5. Learning. Hear again, memory, motivation, and decision stategies are often nested together to accomplish a learning task.

When old-timers in NLP say that most people, on the level of structure have but between three and five basic strategies, this is what we are referring to: the last two strategies often replicate or use pieces of the the first three. Even complicated strategies like creativity, falling in love, negotiation, and national defense will all utilize combinations of the five fundamental strategies.

The TOTE

The general model of a strategy is known as the TOTE (Test-Operation-Test-Exit). This was first formulated in the book Plans and the Structure of Behaviour by George Miller, Eugene Galanter and Karl Pribram published in 1960.

Figure 1: TOTE

The first Test is what sets the strategy into operation. It begins the sequence. You see something you want to buy, or you have a decision to make, or a presentation to give. There are also criteria that are fed forward from this point: the evidence that will let you know that the outcome has been achieved. Without this you can go round and round in the loop and never know when to exit. The TOTE becomes a TOT.

The Operation is what you do to get your outcome, make a picture, talk to yourself, observe, listen, act, etc. The more choices here the more flexibility and the more chance of success. This operation may be very complex and involve other TOTES.

The second Test determines whether you have moved any nearer to your outcome. It is a evaluation of where you are, with where you want to be, using the criteria that have been fed forward form the first Test.

The TOTE is a simple model and it is obvious that there will be TOTEs within TOTEs. An operation within one TOTE may be a whole TOTE at another level, so already we have levels of nested TOTEs.

Tests

A word about tests. In the language of meta-programs, this is where the functions of matching, mis-matching, and comparison come into play, along with the activity of approaching and avoiding. In one of the author’s NLP teaching institute, it has been found very useful to notate how a person is evaluating criteria. In the example of our motivated lad above, he is using comparisons, analogic transformations of brightness and size, to test. This can be notated as a subscript (c/) beside the diagonal line representing a test. Our procrastinator, however is trying to mis-match, and is seeking novelty, and difference. This can be notated with a little (d/). This allows him to keep stalling. Until his representation of the consequences of the unfilled task becomes so different from his imagined diversions that they literally create a mis-match between the pictures in his test. Others might test their pictures, sounds or feelings with a search for sameness (s/) which would indicate someone trying to match criteria. Arrows can also be used on top of the diagonal stroke of the test to indicated approaching or avoiding evaluations.

Exits or Choice Points

Which leads us to the nature of exits, which are perhaps better defined as choice points. This is the last step in a strategy, which occurs just before external action takes place , or another internal strategy is fired off. Exits serve as traffic lights, pressure gauges, clocks or counters in our thought process-- depending on one’s meta-programs. Exits represent the outcome of a test. Is the test satisfactory? Then green-light, exit from the brain and go back out to the external world. Unsatisfactory? Maybe it’s time to stop, and go have a beer? Not sure? Yellow light-- go around another few times, or try another strategy.

How are the messages of yes/no/maybe transferred to our consciousness by the brain? It’s a matter of thresholds again. Intensity. But intensity is produced by several different experiential patterns. One is repetition-- how many times something occurs. So some people take a number of times to go around a given strategy before exiting. Some people favor frequency-- how often something occurs in a given time period.

Others pay attention to duration, or even interval-- how long a time something stays constant, or the period of time between two events.

When do you know someone is a good lover? From one, single, intense experience? Because he or she made love with you three times? Or because they performed three times in a single hour? Or because they did it for three hours straight without stopping? Or is it because it’s been three months since the last time you’ve done it? One of these patterns will appeal to you more than another, according to the meta-program of your decision strategy, and when you exit with conviction (and pleasure).

Micro, Macro and Meta Strategies

The examples of motivation, above, are what you could call a micro-strategy. It is small-chunked, and reproducible. In these articles, we have argued for a strategy model of parallel, interconnected and simultaneous process, expanding and enriching the existing linear, sequential model. Can this be done?

Let us take a look at a macro-strategy, one that includes several nested , concurrent loops. It can be done, after all, and this is a real-life example of how a complex strategy was detected and transferred.

One day one of the authors was in Marilyn Ferguson’s kitchen. Marilyn is the editor/publisher of the Brain/Mind Bulletin, and author of “The Aquarian Conspiracy.” The following creative writing strategy was elicited in under a half an hour.

Figure 2: Marilyn’s Writing Strategy.

When Mariyn first gets an idea, it is as a phrase (Aid). Next she sees a matrix of possibilites, gets a feeling, and talks to herself about the feeling, which produces V/K synesthesia patterns. Then she continues to a feed-forward, “outframe” of all the possibilities inherent in this idea for her friends, society, or even the future of humankind. Simultaneously, she asks herself “What’s it mean for me?”, and begins chunking down ideas. Both parallel processes produce more synthesia representations which eventually fit, or match. Then she begins to see, hear, and write her article, which produces more concurrent images, which produce several kinesthetic experiences of satisfaction, appropriateness, and the intuition to wait for unconscious processing. After this wait, she operates auditively to hear the story she is writing as she exits to operate her keyboard.

The best thing we can say about this model is that it was verified immediately. Just after the elicitation, a 15-year old lad happened to come into the kitchen. The modeller walked the boy through the steps of the strategy, as Marilyn observed. First the boy was asked to get an idea, then to imagine a matrix of visual associations, and so forth, allowing some of the creative process to flow in a parallel manner, juggling three lines of thinking at the same time.

Marilyn listened to what the 15 year-old produced at each junction of the process. At the end of the 15-minute experiment, she heard the final result and stated, “Yes, that’s how I write.”

Planning a presentation

Let’s now take a look at what we call a meta-strategy, or an overall model of a competence that uses several complex strategies in hopes of applying what we know today from neural network computing and neurobiology.

Take the example of planning a presentation. This is an exercise both authors have used in their seminars to aid participants in discovering the concurrent aspects of real-life strategies.

As we mentioned in our last article in this series, we already have a way of thinking about parallel operating strategies from the Emprint Method by Leslie Cameron Bandler, David Gordon, and Michael Lebeau (1985). Briefly, the Emprint Method is a model for investigating the structure of subjective time, and has been created for the transfer of competences rather than as a diagnostic for pathology.

In their work, they chunk down competences into moments of time called operating formats, in which time orientations of past-present-future are modelled out with how criteria is being generated and evaluated in terms of representional systems, personal implication and causality. In brief, it is a magnifying optic for the inner-workings of the test in a decision strategy that generates activity in the external world, producing a verifyable outcome.

They make a distinction between intrinsic behaviours, that the person has acquired in the course of life, and intentional behaviours that the individual has sought out and learned. The latter are more interesting and complicated because they will need all sorts of preliminary strategies such as motivation and planning, The activity is then broken down into a number of operative formats. Each one is a step in itself and can be broken down into smaller pieces and the results fed back at higher levels.

A simplified flow chart of the operating formats, contexts and activities a participant would give us as the steps they followed in preparing a presentation might look like the following, Figure 3.

As soon as we began pointing out the feed-forward and feed-back loops in their process, it was quickly evident that many of their activities, which we have diagrammed here as activity levels A through G, go on simultaneously.

This kind of strategic modelling addresses the fuzzy logic (When is something “good enough?”) of neurocomputing. There are moving thresholds: in some contexts the criteria in this system are applied rigorously and less rigorously in others, depending on the outcome of other TOTEs. Errors are tolerated. This individual is not thinking in terms of right or wrong, yes/no digital signals, but “Maybe”, Perhaps” or, “Leave it for the moment and think of something else.” TOTEs run simultaneously and in parallel, all have to be accessible in a parallel time frame. The TOTEs affect each other systemically. There is information both fed forward and feedback between TOTEs.

Peter Senge’s Systems Dynamics

Outside NLP, People have been using flow charts since the l930s. And Peter Senge at MIT has been elaborating systems dynamics since the late 1970s. In his popular book The Fifth Discipline, Senge talks about using systems thinking in business to create a learning organisation, and his ideas apply just as easily to interacting TOTEs. A human being is among other things, a learning organisation.

For example, Senge describes a systems structure that he calls ‘shifting the burden’. It is the structure of addiction on an individual level. A short term solution is used to correct a deeper long term problem.

Figure 4: Shifting the burden

For example, a presenter who wants to win over his audience might forget some of the points he wants to make during the course of a presentation. He gets over this by applying a symptomatic solution: learning his material more and more carefully. In the end he may be quoting it verbatim, and still not address his fundamental problem: his emotional state while presenting is poor. Because he is not getting in a good emotional state, he relies more and more on learning the material by heart. Applying the ‘learning by heart’ series of TOTEs causes delays in using other ‘getting in a good emotional state’ TOTEs. Also, his audience is complaining that he isn’t “walking his talk, ” which is a side effect of his memorization. His attempted solution has not solved his basic problem: state management. What is more, he will need more and more memorization to keep the status quo.

This introduces three important qualities that we recommend should be added to the NLP strategy technology.

First, the concept of positive feedback loops , or what we would call “feed-forward” cycles. A positive feedback loop is a sequence that reinforces itself. For example, disregarding emotional state creates inappropriate emotions. This worsens the emotional state. The person disregards it even more. A vicious circle is set up. Positive feedback loops can set up virtuous circles too. When our hypothetical presenter attends to her emotional state, she presents better, gets better feedback, becomes more confident, presents even better, gets even better feedback etc.

Secondly, there are negative feedback loops . This is not about criticism, nor does it imply a value judgement. Negative feedback creates a balance or equilibrium. The more you memorise, the less there is left to memorise eventually you have to stop.

Thirdly, there are inevitable time delays that tend to organize our behaviors along the arrow of time. Because feedback is not instantaneous, the connection between cause and effect is not always obvious. Our presenter may go well down the road of memorization before it becomes obvious that it is not solving his problem

Summary

These articles are a first attempt in tackling a complex field. To summarize, the new ideas we suggest for constructing strategies that take into account the modern advances in neurocomputing include:

* Multiple, simultaneous, parallel operating TOTEs

* Moving thresholds that involve critical submodalities and metaprograms.

* Tolerance of errors

* Feed forward and feed back of results

* Systemic interaction between TOTEs

We are not trying to build a model based on neural networks. We are saying that neural networks are better models on which to base the ‘programming’ part of NLP, and both need to take advances in biology and cognitive science into consideration. The brain builds models, so all models of the brain will be inadequate. We can only make our model as code congruent with the way we think the brain works as possible.

We cannot leave this field without talking about emotions. Computers cannot model emotions and they are what give our thinking its richness, scope, unpredictability and creativity. Values and emotions could be the equivalent of the hidden levels in our thinking. So much of our thinking is unconscious anyway, and most of it will always be so. We know many things at different levels. For example, there is a neurological condition called prosopagnosia. Patients with this disorder when shown a picture of a person they knew before their illness, will deny knowing them. Yet a GSR (galvanic skin response) measurement reflects a sympathetic nervous activity that shows an emotional response related to memory. The logic of emotion is different to conscious reason. The language of the heart is different to the language of the head.

The recent ideas in Physics on Chaos theory show that events that look simple on the surface can have unfathomable depths, while events that look complex on the surface may be governed by a simple rule applied recursively. Emergent properties we never suspected suddenly present themselves like faces appearing in a cloud formation. In quantum physics the answer can be yes and no at the same time. Is light made of particles or waves? Both - it depends how you set up your experiment. Set up to catch particles and sure enough you’ll catch some. Set up for waves and there they are.

No surprise then, that we operate by fuzzy logic, where the real world is Aristotle’s nightmare, answers are a definite maybe.

As Gregory Bateson said, “If you want to understand something, think about it in the way that thing thunk.” And perhaps that may or may not be fuzzy logic.

The latest developments in the work of physics are leading in ever new directions. Indeed, Nobel prize winner Ilya Prigogine has just offered mathematical proof of a new formulation of physics: that time is after all, irreversible, and that the very laws of nature are evolving-- and perhaps a function of our interaction with nature.

These articles have not been easy to write, and we have used only a fraction of the material available. Caught in Aladdin’s cave, we have constantly had to neglect fascinating detours. We would appreciate your feedback on any maps or paths we have neglected.

Bibliography

Introducing NLP Joseph O’Connor and John Seymour (Thorsons 1993)

Strategies Wyatt Woodsmall 1988

NLP Volume 1 Robert Dilts, Richard Bandler, John Grinder, Judith DeLozier, (Meta Publications 1980)

The Emprint Method Leslie Cameron Bandler, David Gordon, Michael Lebeau (Future Pace Inc. 1985

The Fifth Discipline Peter Senge (Doubleday 1992)

Joseph O’Connor is an author, trainer and musician who lives in London. His latest book, Training with NLP was published in July.

He can be contacted c/o Lambent Books, 4 Coombe Gardens, New Malden, Surrey, KT3 4AA England

Brian Van der Horst, an NLP Trainer who directs an NLP Institute in Paris, started his professional life as a marine biologist and also writes for “INTELLIGENCE -- The Future of Computing,” an international newsletter specialising in neurocomputing, artificial intelligence and electronic networking. He can be contacted at: Repere, 78 Avenue du General Michel-Bizot 75012 Paris, France.

4979 words